Removing m_hostAssemblyMap and ceremony around it
#61292
Conversation
|
Tagging subscribers to this area: @vitek-karas, @agocke, @VSadov Issue DetailsSince we have only one AppDomain, there is no point in having a per-AppDomain { This allows to remove a few locks, including the one taken during GC stack walks, and coordination with EE suspension. Also removes SHash support for callouts to host in hosted scenario. I do not believe such scenarios exist anymore.
|
|
Fairly simple change - mostly removing code. |
| @@ -278,7 +273,7 @@ class EMPTY_BASES_DECL SHash : public TRAITS | |||
|
|
|||
| // Utility function to allocate new table (does not copy the values into it yet). Returns the size of new table in | |||
| // *pcNewTableSize (finds next prime). | |||
| // Phase 1 of code:Reallocate - it is split to support code:AddPhases. | |||
| // Phase 1 of code:Reallocate | |||
| element_t * AllocateNewTable(count_t requestedSize, count_t * pcNewTableSize); | |||
There was a problem hiding this comment.
Is there still any good reason for Reallocate to be split into 3 methods?
There was a problem hiding this comment.
They are not small/trivial methods and have different responsibilities, so I figured it is ok to leave them as separate.
There was a problem hiding this comment.
There are also other callers - like AllocateNewTable is also called by Grow_OnlyAllocateNewTable.
There was a problem hiding this comment.
Others have multiple callers too.
|
Thanks!! |
| @@ -587,6 +587,8 @@ class DomainAssembly : public DomainFile | |||
| void Allocate(); | |||
| void DeliverSyncEvents(); | |||
| void DeliverAsyncEvents(); | |||
| void RegisterWithHostAssembly(); | |||
| void UnRegisterFromHostAssembly(); | |||
There was a problem hiding this comment.
Thanks! I will do the rename in some other change.
ghost
locked as resolved and limited conversation to collaborators
Since we have only one AppDomain, there is no point in having a per-AppDomain {
HostAssembly-->DomainAssembly} map.Removing this map allows to remove couple locks, including the one taken during GC stack walks, and coordination with EE suspension.
Also removes
SHashsupport for callouts to host in hosted scenario (comments say "hosted", I assume they mean "SQL-hosted").I do not believe such scenarios exist anymore.